When it comes to using GAMS, AMPL, or LINGO what are the pluses and
minuses in-between these three packages?
Expert non-commercial reply would be greatly appreciated indeed.
Mike
> Hi eveyone,
>
> When it comes to using GAMS, AMPL, or LINGO what are the pluses and
> minuses in-between these three packages?
GAMS:
Large customer base, old, somewhat limited in its modelling capabilities
AMPL:
Much smaller customer base, very quick response on bug fixes,
development of new features can take a long time, limitless
expressivity.
LINGO:
?
See MPL (www.maximalsoftware.com) and its OptiMax component library.
MPL is a robust algebraic modeling language. With OptiMax you can
embed the formulation in stand-alone applications. So you formulate
with MPL and when you have it right you move the tuned formulation to
OptiMax. Seamless links to most solvers. MPL passed GAMS by sometime
ago. I have no financial interest, just use the product. BTW, last
used LINGO 3 years ago. Very poor IP/MIP performance versus my MPL
supported solver XA.
Good Luck,
Steve
Bob
"Michael" <mas_i...@yahoo.com> wrote in message
news:118f1485.03101...@posting.google.com...
Hi again everone,
What about commercial optimization software packages that allow you to
use, for example, Genetic Algorithm with/or Branch-and-Bound? Also,
solution sensitivity analysis is very important component to our
application. Which product does a good job on the last?
Thanks,
Mike
>
> What about commercial optimization software packages that allow you to
> use, for example, Genetic Algorithm with/or Branch-and-Bound? Also,
> solution sensitivity analysis is very important component to our
> application. Which product does a good job on the last?
This does not depend on the modeling language, but on the solver
you call with it...
Arnold Neumaier
Arnold,
Yes, that what I did mean. This has nothing to do with any modeler. We
are talking about solvers here; good ones. Many commercial solvers do
use CPLEX which uses, for say, branch-and-bound. However, I don't know
if there is a commercial solver that allow you to combine
branch-and-bound and GA for example?
Thanks Arnold,
Mike
> We are talking about solvers here; good ones. Many commercial solvers do
> use CPLEX which uses, for say, branch-and-bound. However, I don't know
> if there is a commercial solver that allow you to combine
> branch-and-bound and GA for example?
Are you really talking here about good solvers?
GA is quite slow as a general purpose solver. So it does not
make sense to combine it with branch-and-bound. In a
branch-and-bound scheme you want to use only quick heuristics;
otherwise the combination will be poor.
Arnold Neumaier
AMPL has an open interface, so any solver can be hooked. A lot more NLP
solvers already have AMPL hook code than any other modeling language.
Plus there are hooks to MIP solvers such as CPLEX. I know that hooks
exist for at least CPLEX, MINOS, SNOPT, LANCELOT, IPOPT, KNITRO, LOQO,
filterSQP, MINLPBB, DONLP2, PENNLP. Solvers need to be purchased
separately and hooked in.
MPL I don't know much about, but I think there is a somewhat more
limited solver list than AMPL or GAMS. Windows only.
OPL is the ILOG language which is only compatible with CPLEX and ILOG
Solver. It does allow for MIP and constraint programming hybrid models.
Mosel is the Dash language supporting XpressMP (LP/MIP/QP) and XpressSLP
(NLP,MINLP). It will probably also support the not yet released
XpressSP (stochastic programming) and XpressCP (constraint programming)
solvers.
AIMMS from Paragon has a nice environment, and has solvers including
CPLEX, XpressMP, XA, CONOPT, MINOS, SNOPT and an in-house
outer-approximation code called AOA for MINLPs. Windows only.
LINGO has LINDO for LP, I'm not sure what they use for MIP or QP, and
they have CONOPT for NLP. LINGO also has a home grown global optimizer
supposedly based on the same approach as BARON. ?Windows only?